update: 2022.01.19 RL is not used in real world. so my assumption and prediction about tesla semi truck is completely wrong. https://youtu.be/zR11FLZ-O9M?t=3650
below is what i 'think' what might happen.
due to changes in states, the random variables have to be increased. the changes in states include cameras and sensors in a different angle or part of the car, the mass of the car, brakes responses ... and the current learned model(reinforcement learned model, not car model) will help the new model to be trained, but the margin of 'errors' or crashes is the only answer to make the new model learn properly.
Otherwise, the autonomous driving experience won't be the same as other tesla models.
Now the mass of the crashes will be much higher, so it will result in more severe deaths, thus leading to social issues to force tesla to open their codes, specifically random variables.
In the end, tesla will release 'safe' version of the autonomous drive with no random variable. Then, drivers can choose whether they want to be lab mouses. Tesla will no longer be able to use free labor and will have to pay drivers who choose to be lab mouses.
Tesla will strategically raise this issue when time is ready. the time is determined by whether the data is completely collected by the tesla side, but not other manufacturers.
